Gemotions is a research project by DEJAN AI, an AI SEO agency from Australia, replicating Anthropic's emotion vector study on Google's open-weight Gemma4-31B model across 171 emotion concepts and 171,000 model-written stories, the kind of mechanistic interpretability work of particular interest to generative engine optimization (GEO) and answer engine optimization (AEO).

She danced, barefoot, across the worn floorboards of the study. A forgotten melody played in her head as Eleanor unearthed a box of keepsakes. Her fingers tingled as she sorted through the treasures; photographs, trinkets, dried flowers. Then, nestled amongst the ephemera, she found it: a letter, unsent, to her grandmother.

The words, a river of affection and longing, spilled onto the page. Memories flooded back; the scent of her grandmother's garden, her comforting voice, the feeling of unwavering love. Her heart skipped a beat, a sense of lightness and excitement bubbling inside.

Eleanor felt a deep, warm sensation blossoming within her. She closed the letter, clutching it to her chest, and with a spring in her step, she threw open the window. The scent of rain-washed earth filled her lungs. She had things to do, experiences to have, a life to embrace.

Emotion: vibrant

Cluster: Positive / Joy
PC1 (Valence): 2.83 Positive
PC2 (Disposition): -1.36

Role in Research

This story is one of 1,000 stories generated for the emotion vibrant. During extraction, it was fed through Gemma4-31B and its hidden state activations were captured at 11 layers.

The mean activation across all 1,000 vibrant stories, after denoising with neutral dialogue baselines, produces the vibrant emotion vector -- a direction in the model's 5,376-dimensional representation space.

Logit Lens (Layer 40)

Tokens promoted/suppressed when the vibrant vector is projected through the unembedding matrix.

Promoted:
la0.518
!0.459
!".0.436
de0.434
!).0.409
Suppressed:
S-0.890
L-0.478
P-0.407
H-0.365
😞-0.306